memory-auditlisted

# memory-audit Use `memoryvault.memory_health` to get a summary of vault health. ## What you get ```json { "n_memories": 470, "pct_with_entities": 0.97, "n_entities_in_use": 158, "useful_entities": 94, "dead_wikilinks": 0, "orphan_entities": 153 } ``` ## How to interpret - `pct_with_entities` should be ≥0.95. If lower: 5%+ of memories are isolated from the entity graph — they won't surface via graph walk. - `dead_wikilinks` should be 0. If >0: ingestion wrote `[[Name]]` references to entities that don't exist. Run `memory heal --apply` to fix. - `orphan_entities` is acceptable up to ~30% of `n_entities_in_use`. Higher means lots of dead entity files (entities exist but no memory mentions them). - `useful_entities` should be ≥40% of `n_entities_in_use`. These are entities with 2-20 memories — the "discriminating" range for graph walk. ## When to escalate If any of these are red, surface the issue to the user and suggest: - dead_wikilinks > 0 → "Run `memory heal --apply`" - pct_with_entities < 0.9 → "X% of your memories have no entity wikilinks; consider re-running ingest on those" - useful_entities / n_entities ratio < 0.4 → "Most of your entities are either singletons or hubs; the graph isn't doing much work"